Diagnostic agents for many diseases are based on the ability of antibody reagents to bind tightly and specifically to sugars on cell surfaces that are markers for the disease. Knowledge of the 3D structure of these interactions helps to optimize the reagent and aids in confirming its specificity, however traditional experimental methods are often poorly suited to working with large flexible sugars. Consequently, alternative high- throughput techniques, such as screening the protein or antibody reagent against immobilized arrays of glycans have become popular for identifying the binding partners. Array screening does not however provide any information on the 3D structure, and can be prone to false negative or non-specific results. The principle aim of this proposal is to develop a robust high-throughput computational platform for predicting the 3D structure of carbohydrate-antibody (or protein in general) complexes, which will integrate data from glycan microarray screening to define the origin of the observed binding specificity, with the ultimate goal of aiding in the development of high-specificity diagnostic agents. This work will proceed through the following stages 1) The development of automated tools for docking the minimal glycan binding motifs, superimposing each of the glycans from the experimental screening data set that share the motif, and determining experimentally- consistent 3D poses 2) The development of a computational scoring function that is tuned for use with carbohydrates, and that incorporates the potential energy functions from the GLYCAM carbohydrate force field 3) The creation of an annual competition to predict glycan-protein 3D structures involving the theoretical and experimental glycoscience communities 4) The creation of a web-based set of tools that allows users to upload their protein structure and glycan array data and perform Carbohydrate Threading to generate an experimentally-consistent 3D model for their carbohydrate-protein complex This work is highly innovative in integrating high-throughput computational and experimental techniques to generate experimentally-consistent 3D models for otherwise hard to characterize molecular complexes. Notably the approach builds on the strengths of each method and concurrently aids in identifying the errors commonly found in array screening and theoretical ligand docking. PUBLIC HEALTH RELEVANCE: The development of highly specific diagnostic agents would benefit greatly from the ability to study their 3D structures;however traditional experimental methods are often poorly suited to working with some typical disease-associated markers, such as large flexible sugars. Consequently, high-throughput experimental techniques, such as screening the diagnostic reagent against immobilized arrays of complex sugars (glycans) have become popular for identifying the binding partners;but array screening alone does not provide any 3D information. The principle aim of this proposal is to develop and integrate theoretical 3D structure prediction methods with data from glycan array screening to generate computational tools to aid in the development of high-specificity diagnostic agents.